Does recollecting IPV experiences change women’s preferences?

Our research focuses on poor married women from Bangladesh with a high IPV rate of 72.6%.
To examine the impact of recollection of IPV experiences on women’s preferences, we will remind women about their IPV experiences inspired by existing studies that use priming of recent trauma (Lerner and Keltner, 2001; Lerner et al., 2003; Callen et al., 2014).

In particular, the respondents will be asked several questions about their emotional, physical and emotional IPV experiences, after which a short video will be shown to them, which demonstrates the abuse from the husband in front of the children. The video was developed by the WE CAN Bangladesh, one of the local organisations that work to make women aware of the violence and available support to fight against violence in Bangladesh (https://www.youtube.com/watch?v=pGvY3e6t-vE&t=1s&ab_channel=wecancampaign).

The IPV-related questions and the video are jointly referred to as IPV priming. The respondents who will be primed before the outcome module are referred to as treatment and the others are referred to as control.

2023-06-11

2024-03-31

Risk preferences, time preferences, social preferences and intra-household decision-making preferences

Following the Global Preference Survey (GPS) analysis, we will focus on the following measures.
Risk preferences
To measure risk aversion, we will ask 31 questions for quantitative and 1 question for qualitative measures (i.e., self-reporting risk-taking behaviour) (Falk et al., 2018). The quantitative measure consists of 31 questions with two options, a 50% chance of receiving the full amount or a sure payment. For each question, the sure payment varies. The qualitative measure consists of self-assessed risk-taking behaviour on a scale of 0-10, with 10 referring to the strongest willingness.
Time preferences
To measure impatience, we will ask 31 questions for quantitative and 1 question for qualitative measures (i.e., willingness to wait) (Falk et al., 2018). The quantitative measure consists of 31 questions with two options, a payment today or a larger payment in 12 months. For each item, the future payment varies. The qualitative measure consists of the self-assessed patience level on a scale of 0-10, with 10 referring to the highest patience.

Positive reciprocity
To measure positive reciprocity, we will ask two questions as follows: (i) preference (in monetary form) for offering a gift in exchange for help and (ii) self-assessment about willingness to return a favour. Both questions will be assessed on a scale of 0-10, with 10 referring to the strongest willingness.
Negative reciprocity
To measure negative reciprocity, we will use three questions to elicit their willingness (measured on a scale of 0-10, with 10 referring to the strongest willingness) to (i) take revenge, (ii) punish themselves for unfair behaviour, and (iii) punish someone for unfair behaviour.

Trust
We will ask one question to measure trust level by asking to what degree they believe people have only the best intentions on a scale of 0-10, with 10 referring to the highest belief.

Altruism
We will use two questions on the donation: (i) how much of an unexpected monetary gain they want to donate and (ii) self-assessment of willingness to donate on a scale of 0-10, with 10 referring to the strongest willingness.

We will construct indices using these measures following Falk et al. (2018). In particular, we will have three indices: risk, time and social (including reciprocities, trust and altruism). They use weight for each question of an index. The sum of the weights needs to be one for each index. Therefore, if there are multiple items under one weight (i.e., multiple draws to measure risk aversion), the weight is equally distributed among the items.
The steps to construct an index are as follows; i. compute the z-scores of each survey item at the individual level, (ii) weigh these z-scores using the weights (reported in column 6 of Table A1 in the pre-analysis plan), and (iii) sum the weighted z-scores for each index. For example, risk-aversion index= (0.01525 X z-score of draw 1)+ (0.01525 X z-score of draw 2)+……+ (0.01525 X z-score of draw 31)+ (0.527 X z-score of self-assessed item).

Our list of outcomes also includes intra-household decision-making preferences. To elicit these preferences, we will adapt and expand the standard module of DHS on women’s participation in household decision-making (NIPORT and ICF, 2020). For example, while the DHS bundles all household major decisions together within one question, we will separate these decisions into more specific domains. In particular, we will ask women whether they are willing to make decisions regarding (i) household daily expenses; (ii) food consumption; (iii) how their individual earnings will be spent, (iv) household saving decisions (where to save/how to spend savings) (v) children’s education; (vi) marriage of children; (vii) medical treatment of household member; (viii) Visits to family or relatives; and (ix) own health care. Their response will be coded on a scale of 0-10, with 10 referring to a stronger willingness.

Information on household composition, occupation, climate change resilience, etc., will be collected from all women, and they will be asked about their emotional, physical and sexual IPV experiences in their lifetime and last 12 months; however, each individual will be asked about one of the outcome (preference) modules. Particularly, we will have six groups (G1, G2, and so on) to implement six sets of survey instruments. The groups are as follows.

G1: The group will be asked questions about their IPV experiences, after which the video will be shown. Finally, they will be asked the questions to elicit their risk preferences.
G2: The group will be asked questions to elicit their risk preferences, after which they will be asked questions about their IPV experiences and shown the video.
G3: The group will be asked questions about their IPV experiences, after which the video will be shown. Finally, they will be asked the questions to elicit their time preferences.
G4: The group will be asked questions to elicit their time preferences, after which they will be asked questions about their IPV experiences and shown the video.
G5: The group will be asked questions about their IPV experiences, after which the video will be shown. Finally, they will be asked questions to elicit their intra-household and social preferences.
G6: The group will be asked questions to elicit their intra-household and social preferences, after which they will be asked questions about their IPV experiences and shown the video.
We will randomly assign these six sets of survey instruments at the individual level, giving us more power to detect the treatment (i.e., IPV priming) effect.

BIGD will randomly select the respondents from the list of eligible participants for the programmes. Then, we will randomly and equally divide these selected respondents into the proposed six groups using STATA.

Individual level

No

N/A

We have three sets of outcomes, and each set has individual treatment and control groups. Treatment and control groups of each set will have equal observations. Since the first phase of the survey will cover about 1200 women, each group (G1-G6) will have 200 samples. In the second phase, the survey will cover 800 women, and we will divide these samples into effective groups based on the findings of the first phase. For example, if we find promising impacts on risk preferences only, we will randomly divide 800 samples into G1 and G2.

We have three sets of outcomes, and each set has individual treatment and control groups. Treatment and control groups of each set will have equal observations.

Since the first phase of the survey will cover about 1200 women, each group (G1-G6) will have 200 samples. In the second phase, the survey will cover 800 women, and we will divide these samples into effective groups based on the findings of the first phase. For example, if we find promising impacts on risk preferences only, we will randomly divide 800 respondents into G1 and G2.

For effect size calculation, we use the baseline mean and standard deviation (SD) for risk-taking attitudes (on a scale of 0-10, with 10 referring to taking the highest risk), social preferences (on a scale of 0-10, with 10 referring to the highest trust in other people), and time discounting (percentage) from two studies (Callen et al., 2014; Voors et al., 2012). Using a 95% confidence level, a sample size of 400 from each group gives a minimum detectable effect size (MDE) of 0.20 points at 80% power for these outcomes.